/supervoxel-splits response includes duplicate entries

[stuarteberg]

The /supervoxel-splits endpoint includes many duplicates. Here are two examples:

In [143]: dvid_splits_df.query('mutid == 1002224862')
Out[143]:
                                uuid       mutid         old      remain       split     type
43  ef1da1a01c4941b6876ab77502843d0c  1002224862  5813075645  5813075647  5813075646  unknown
44  ef1da1a01c4941b6876ab77502843d0c  1002224862  5813075645  5813075647  5813075646  unknown

In [144]: dvid_splits_df.query('mutid == 1005819021')
Out[144]:
                                   uuid       mutid         old      remain       split     type
34651  3c281c385e064ed1b6ab93a9456a04eb  1005819021  2349519569  5901205938  5901205937  unknown
34652  3c281c385e064ed1b6ab93a9456a04eb  1005819021  2349519569  5901205938  5901205937  unknown

Some notes:

• The /supervoxel-splits endpoint includes the results from both /split-supervoxel and /split commands, but apparently the duplicates come from ONLY /split-supervoxel commands. No /split commands resulted in a duplicate entry.

• Out of the 13174 /supervoxel-split commands that have been issued on the production master branch**, 12861 resulted in duplicate entries. (And 313 did not.)

• On the current production node (36b0), there have been 288 /split-supervoxel actions so far. However, NONE of these are duplicated in the /supervoxel-splits output. In other words, when I said above that 313 of the splits were NOT duplicated, most of them are from the current node. (But not all of them.)

** Not including the root node 2884, which isn't represented in our kafka logs.